LangChain Tutorial (JS) #3: Output Parsers

#openai #langchain #langchainjs

We can convert the AI-responses into complex structures, like CSV, JSON, Arrays and much more using LangChain Output Parsers.

☕ Buy me a coffee:

📑 Useful Links:

💬 Chat with Like-Minded Individuals on Discord:

🧠 I can build your chatbots for you!

🕒 TIMESTAMPS:
00:00 - Introduction to Output Parsers
00:52 - Project Setup
01:14 - Output Parser Types
02:00 - String Output Parser
02:38 - Quick note on LCEL
03:11 - List Output Parser
03:21 - Refactor Code
06:45 - Structured Output Parser
11:25 - Zod Stuctured Output Parser

[ramafa]

Excellent tut. Other youtubers made this topic so complicated but you explained it so well!! You are underrated

[eivanmtzleal]

Great! I'll be keeping an eye on your channel! There is very little "LangChain" content in JS, thank you very much

[cyborgmetropolis]

This series took me pretty far! Invaluable!

[RABWA]

this is really amazing content, the way of explanation is quite clear, thanks a lot

[jandramila]

Very useful, It even helped me understand some of the python code that's out there using this same feature.

Props to you! Subscribed and liked your series

[salmantanvir]

you just earned a sub my friend, great series!!

[zhitoobit]

please explain long chat history with redis. thank you

[michal-xp]

Hey great video. How can I get amount of tokens used when using parsers?

[abirdutta]

Hey StrcutredOutputParser giving parse erros sometimes as LLM not following the format instructions everytime.

[yoelfdz]

Is it possible to stream api using an output parser? Struggling to make that work

[abirdutta]

Hey can you explain where to add the output parser if I wanna use loadQAStuffChain from lanchain js?

[tebogo]

For the zod out try this -

const callZodOutputParser = async () => {
const prompt =
`Extract information from the phrase and the answer,
Formatting Instructions: {format_instructions}
Phrase: {phrase}`
);

const outputParser =
z.object({
recipe: z.string().describe("name of recipe"),
ingredients: z.array(z.string()).describe("ingredients"),
})
);
const chain =
return await chain.invoke({
phrase: "Can you tell me the ingredients to a Spaghetti Bolognese",
format_instructions: outputParser.getFormatInstructions(),
});
};